A long-awaited launch
On November 10, OpenAI opened Businesses in ChatGPT. My first thought was: finally. My second was to wonder what it says about apartments.
OpenAI says about 1.2 billion people use ChatGPT every week. Until now, a business had no direct way to speak to them there. Now it can claim a verified profile and publish its own prices, and in some categories it can put up an agent that answers questions on its behalf. Meta's Muse recently showed where personal AI agents are headed. This shows where the businesses those agents deal with will be found. It is fitting that the product with the largest audience moved first.
I have seen this before
Before VALIS, I spent more than a decade building a media and ratings business for boutique hotels. Every few years a platform changed the rules: booking sites opened guest reviews, and later social apps opened business accounts. The hotels that moved early gained years of advantage. Once everyone was on the platform, being there no longer set anyone apart. While the window was open, though, the difference between early and late was large.
That window has now opened for apartments.
What we looked at in Charlotte
The day after the launch, my team took a first look at what multifamily looks like inside ChatGPT. We searched ChatGPT's business directory for 117 randomly selected apartment communities in the Charlotte area, and I asked ChatGPT the questions a renter would ask. Four observations stood out.
What we found
1. The profiles already exist. 115 of the 117 communities already had a directory entry, created automatically. For owners, the question has moved on. Their properties are already there. What matters now is who defines what the profile says.
2. About one in five is inconsistent. 25 of the 117 had a duplicate entry, a different name, or an address that did not match the community's own. Ellison Mallard Creek also appears as Optima Mallard Creek. Rise Sunridge shares its address with Arrowood Crossing, the property's name before a 2025 sale. Century University City shares its address with a furnished rental company. Several communities have separate entries for a garage, a gym or a single building.
These details reach renters. Asked about Optima Mallard Creek, ChatGPT recognized it as Ellison but cited the rating from the old entry. Asked about Ello House, now Camden LoSo, it gave one set of prices under the old name and a different set under the new one.
3. ChatGPT repeats what a property publishes. None of the 13 answers I received showed a verified profile. Every answer was assembled from the open web. Where a community publishes its total monthly price and lists each fee, ChatGPT passed that information on in full. For Century University City, it listed all four mandatory monthly fees, down to a $3.75 billing charge, and gave the same list when I asked again in a separate chat. Where a community publishes only base rent, ChatGPT gave base rent and suggested contacting the leasing office.
When I asked, without naming any property, for apartments near UNC Charlotte with transparent fees, ChatGPT ranked communities by whether they publish the full monthly price with every fee itemized. The property it placed first changed between my two attempts. The basis for the ranking did not.
4. Answers vary by person. ChatGPT tailors its answers to the person asking. In a new chat it applied a budget and commute I had mentioned in an earlier conversation. In a free account, four of the ten answers about a specific community also carried an advertisement, one of them from another operator.
What owners can do now
For owners, asset managers and marketing leaders, I believe this is the time to claim profiles and correct them, before AI assembles a property's description from old names and other businesses' listings. The process is not difficult, but it involves people who rarely work together. When I claimed the VALIS profile, ChatGPT asked for proof of control over the company's web domain, a photo ID for identity verification, and the company's legal entity details. In most companies those sit with IT or a website vendor, with the person doing the work, and with finance or leadership. Our own profile is still under review. We are among the first companies to go through the process, and I will share what we learn.
The second stage
Looking further ahead, I expect a second stage. Once most operators have claimed their profiles, every AI answer will carry the operator's own account of the facts, and presence alone will not distinguish anyone. The question that matters then is whether what AI tells a renter matches what the renter actually pays. Answering it will require someone outside the transaction to check.
A note on method
This was one market, observed on one day, and I asked the questions myself from one account. The directory does not show which profiles have been claimed. Prices cited here are ChatGPT's figures, not prices we have verified. Two of the discrepancies we first recorded turned out to be errors in our own records, and ChatGPT had both right; we have corrected them. The full data and methodology are in the accompanying VALIS research note. If any detail about a property named here is inaccurate, please let me know and we will correct it.
Next, I will compare what AI tells renters about mandatory fees with what the properties themselves publish.
Jinbo Chen is the founder and president of VALIS, which independently checks what AI tells renters about apartment communities.